Combine all NVMs on AWS instance (e.g. i3) as raid0 and mount as data

nvmraid.sh

# https://docs.aws.amazon.com/AWSEC2/latest/UserGuide/raid-config.html
nvmes=$(sudo lsblk | awk '/^nvme/ {printf "/dev/%s ", $1}')
sudo mdadm --create --verbose /dev/md0 --level=0 --name=my_raid --raid-devices=$(echo $nvmes | wc -w) $nvmes
sleep 10# crutch
sudo mkfs.ext4 -L my_raid /dev/md0
sudo mdadm --detail --scan | sudo tee -a /etc/mdadm.conf
sudo dracut -H -f /boot/initramfs-$(uname -r).img $(uname -r)
sudo mkdir /data
sudo mount LABEL=my_raid /data
sudo chown ec2-user:ec2-user /data/

I could not find the above resource for the full Python version, so I wrote one. Pasted below.

#!/usr/bin/python3

import json
import os
import shutil
import subprocess
import time

RAID_LABEL = "local_raid"

def mount(name = "LABEL={}".format(RAID_LABEL), path = "/mnt/local"):
    os.makedirs(path, mode=0o777)
    subprocess.run(["mount", name, path], check = True)

def main():
    devs = []
    res = subprocess.check_output(["lsblk", "-J"], stderr=subprocess.STDOUT)
    jd = json.loads(res.decode())
    for bd in jd["blockdevices"]:
        if not bd.get("mountpoint") and not "children" in bd:
            devs.append("/dev/{}".format(bd["name"]))
    if not devs:
        return
    if len(devs) == 1:
        mount(name = devs[0])
        return

    if not shutil.which("mdadm"):
        subprocess.run(["yum", "install", "-y", "mdadm"], check = True)

    cmd = ["mdadm", "--create", "/dev/md0", "--level=0", "--name=%s" % RAID_LABEL, "--raid-devices=%d" % len(devs)]
    cmd.extend(devs)
    subprocess.run(cmd, check = True)

    while True:
        with open("/proc/mdstat", "r") as f:
            mdstat = f.read()
        if "active raid0" in mdstat:
            break
        time.sleep(1)

    subprocess.run(["mkfs.ext4", "-L", RAID_LABEL, "/dev/md0"], check = True)
    with open("/etc/mdadm.conf", "w") as f:
        subprocess.run(["mdadm", "--detail", "--scan"], stdout = f, check = True)

    res = subprocess.run(["uname", "-r"], capture_output = True)
    uname = res.stdout.decode('utf-8').strip()
    subprocess.run(["dracut", "-H", "-f", "/boot/initramfs-%s.img" % uname, uname], check = True)

    mount()

main()

Note that if using the above with AWS Batch, you will need to wrap the above in MIME.